The invention relates in general to audio signal processing and in particular to a method and device for noise suppression in audio signal data processing.
In the processing of signals (e.g., audio signals) in a data processing device, a problem often arises involving a quantization noise in an algorithm due to a limited word width, for example, a word width of 18 or 24-bits. FIG. 6 illustrates a prior art data processing device 10 having a data input 12 for receiving data, a storage device 14 with storage locations for saving the received data as storage data, a data output 16 for providing output data from the storage device 14, and a data processor 18 for processing the storage data. The device 10 may also include a scaling arrangement for scaling the received data with an entry scaling value. The data processor 18 may execute an algorithm that acts on the storage data. The data processing device 10 can be a simple delay element for transferring data values of a sequence of data with a time delay.
The limited word width of 18 or 24-bits commonly used with such a data processing device 10 often produces quantization noise in an algorithm acting on the storage data.
To reduce the noise, a data processing device may utilize a relatively larger word width. When using floating point numbers, each word includes a fixed-point portion and an exponent portion, which may require special hardware and a corresponding word width, since both the exponent portion and the fixed-point portion are operated on and saved.
The influence of quantization noise may be problematical when using noise-amplifying algorithms (e.g., digital filters) with low input levels. A scaling of the received input data or a corresponding back-scaling of output data from the storage device for better selection of the available word width can result in signal distortion. A time discrepancy may exist between the receipt of unscaled data values or those scaled with a different scaling value and put into the storage device, on the one hand, and the output of data values previously put into the storage device without scaling or storage data changed in scale for provision as output data. This time discrepancy between the input and output scaling values may be a problem.
What is needed is a method and a device for noise suppression in a data processing system that are relatively easy to implement and utilize a relatively limited word width.